The default model is the answer AI Router falls back to when it makes no decision for a request. It does not restrict routing: AI Router still chooses among your enabled models, whatever the default is.
Choose your default model in the configuration step during onboarding. The setting is stored for your team and applied server-side to every request that does not name a model. You can change it at any time: open your team in the dashboard and go to Settings → Routing, where the routing policy (default model and preferences) is saved as one unit.
You can choose any model from the dashboard Models page, or choose auto to use AI Router's built-in default, gpt-6-luna at medium reasoning effort. Outside full privacy mode, a default that names a removed model is replaced by the built-in default as well.
Where the default model applies:
model_routing: false, used by the SDKs) returns the default model when AI Router makes no decision. If your team has connected provider keys, such a request can fail with 503 instead.503.candidates list or an error instead.Set the model parameter when a specific request needs a different default model from your account default. To run a request on one specific model, send that model as the only entry of models instead (see Choosing Models). For example:
Using OpenAI:
Using langchain:
You can also set model to auto to use the team default for that request.
Per-request parameters override your account configuration. When a request sends routing parameters, the AI Router uses those values for that request instead of your account configuration. Requests that hard-code routing parameters will not pick up later changes you make in the dashboard.
...
client.chat.completions.create(
model='gpt-6-luna_low',
messages=[{"role": "user", "content": input}]
)
llm = ChatOpenAI(
model='gpt-6-luna_low',
base_url='https://api.airouter.io',
)